Pursuing Right Prospects Using Predictive Analytics
A Case at Sodexo



Vikas Mittal & Hari Sridhar

14 November, 2021

Background

Data

## [1] 3746

Old Provider vs New Provider

Old Provider vs New Provider

Old Provider vs New Provider

Old Provider vs New Provider - By Opportunity Global Region

Old Provider vs New Provider - By Service Level 1

Old Provider vs New Provider - By Strategic Segment L1

Old Provider vs New Provider - By Regional Strategic Account

Regression Analysis - Logistic Regression

Logistic Regression Model Predicting Sodexo Winning the Bid
Sodexo Winning the Bid
Full Data
Old ProviderAramark -0.967***
(0.325)
Old ProviderCompass -0.220
(0.154)
Old ProviderSodexo 1.163***
(0.091)
Opportunity Global RegionAMENA -13.105
(309.686)
Opportunity Global RegionASIA PACIFIC 0.267**
(0.125)
Opportunity Global RegionBRAZIL -0.510***
(0.140)
Opportunity Global RegionCONTINENTAL EUROPE 0.588***
(0.181)
Opportunity Global RegionLATAM -0.146
(0.142)
Opportunity Global RegionNO REGION -2.597**
(1.021)
Opportunity Global RegionUK I
(0.149)
Service Level 1Food 0.062
(0.145)
Service Level 1Integrated services -0.195
(0.142)
Service Level 1Soft FM 0.291**
(0.140)
Strategic Segment L1Corporate Services 0.347
(0.493)
Strategic Segment L1Energy -0.086
(0.100)
Strategic Segment L1Government Services 0.508
(0.603)
Strategic Segment L1Healthcare -14.210
(882.743)
Strategic Segment L1No Segment -13.030
(509.088)
Strategic Segment L1Schools 15.173
(882.743)
Regional Strategic Account1 0.267
(0.209)
year_greater_than_20161 0.538***
(0.136)
Revenue (Annual) (converted) in Mill Euros -0.051***
(0.016)
Margin (GP/BR) -0.006***
(0.002)
Constant -1.390***
(0.193)
N 3,746
Log Likelihood -2,066.078
Akaike Inf. Crit. 4,180.155
Notes: ***Significant at the 1 percent level.
**Significant at the 5 percent level.
*Significant at the 10 percent level.
Base Case - Old Provider = ‘Other’; Opportunity Global Regian = ‘NORAM’; Service Level 1 = ‘Hard FM’; Strategic Segment L1 =‘Mining’

Regression Analysis - Multinomial Logistic Regression

Multinimial Logistic Regression Model Predicting New Provider
Aramark Compass Sodexo
Old ProviderAramark 3.628*** 1.197*** -0.453
(0.448) (0.439) (0.340)
Old ProviderCompass 1.786*** 2.777*** 0.259
(0.546) (0.189) (0.164)
Old ProviderSodexo 1.708*** 2.647*** 1.597***
(0.486) (0.166) (0.100)
Opportunity Global RegionAMENA -4.994*** -11.177*** -12.260***
(0.0001) (0.00000) (0.00000)
Opportunity Global RegionASIA PACIFIC -6.560*** 0.513** 0.334***
(0.003) (0.229) (0.129)
Opportunity Global RegionBRAZIL -6.518*** -0.063 -0.509***
(0.002) (0.260) (0.143)
Opportunity Global RegionCONTINENTAL EUROPE 0.112 0.427 0.662***
(0.884) (0.297) (0.191)
Opportunity Global RegionLATAM 0.636 -0.183 -0.158
(0.618) (0.288) (0.147)
Opportunity Global RegionNO REGION -9.557*** -12.265*** -2.656***
(0.00000) (0.00000) (1.024)
Opportunity Global RegionUK I 1.603*** 0.091
(0.583) (0.291) (0.156)
Service Level 1Food 0.671 0.306 0.122
(0.807) (0.275) (0.149)
Service Level 1Integrated services 0.722 0.462* -0.119
(0.868) (0.281) (0.146)
Service Level 1Soft FM 0.580 0.042 0.300**
(0.814) (0.288) (0.143)
Strategic Segment L1Corporate Services -6.818*** -11.218*** 0.341
(0.00002) (0.00000) (0.500)
Strategic Segment L1Energy -1.105** 0.385** -0.055
(0.523) (0.192) (0.103)
Strategic Segment L1Government Services -8.565*** -10.241*** 0.474
(0.00001) (0.00000) (0.616)
Strategic Segment L1Healthcare -0.016*** -3.565*** -11.957***
(0.00001) (0.00001) (0.00000)
Strategic Segment L1No Segment -4.778*** -7.718*** -11.665***
(0.0002) (0.00000) (0.00000)
Strategic Segment L1Schools -0.007*** -0.362*** 13.849***
(0.000) (0.000) (0.000)
Regional Strategic Account1 -6.079*** 0.296 0.270
(0.0002) (0.396) (0.215)
year_greater_than_20161 -0.516 -0.843*** 0.370***
(0.564) (0.217) (0.141)
Revenue (Annual) (converted) in Mill Euros 0.031 0.056*** -0.043***
(0.035) (0.015) (0.016)
Margin (GP/BR) -0.012 0.003 -0.005***
(0.010) (0.003) (0.002)
Constant -4.370*** -3.389*** -1.307***
(0.924) (0.358) (0.198)
Akaike Inf. Crit. 6,008.627 6,008.627 6,008.627
Notes: ***Significant at the 1 percent level.
**Significant at the 5 percent level.
*Significant at the 10 percent level.
Base Case - Old Provider = ‘Other’; Opportunity Global Regian = ‘NORAM’; Service Level 1 = ‘Hard FM’; Strategic Segment L1 =‘Mining’

Using Machine Learning to Improve Predictive Accuracy

Random Forest

rf_grid <- expand.grid(mtry = seq(2, 8, by = 1),
                      splitrule = c("gini", "extratrees"),
                      n_trees    = c(150,175,200,225,250,275,300,350,
                                     400,450,500,550,600,650,700,750,800,
                                     850,900,950,1000),
                      min.node.size = c(1, 3, 5,10),
                      sampe_size = c(0.5,0.6,0.8))

Random Forest

Application

Link to the App to Predict Win Probability